摘要
结合独立分量分析与支持向量机,提出一种基于特征优化算法的磁共振脑组织分割方法。首先,从图像中提取出灰度和纹理特征构成原始特征集;然后,利用独立分量分析技术对所提取的原始图像特征进行优化处理,提取其中的独立分量构成特征子集;最后,把训练样本与待分类样本都映射到特征子集所张成的独立空间中,利用特征子集对支持向量机分类器进行训练并对脑组织进行分类。实验结果表明,采用本研究的分割方法可以获得比其他相关方法更好的脑组织分割结果。
With combinaton of independent component analysis （ICA） and support vector machine （SVM）, a method for MR brain image segmentation was presented based on feature optimization algorithm. The gray and texture features were extracted from MR brain images, which formed the original features set. Then ICA was used to process the original features set optimistically so that the independent componentcan can be extracted from the original features set to construct a features subset. Finally the samples both trained and unclassified were mapped to the independent space which was constructed by the features subset. The SVM classifier was trained using the features subset to classify the brain tissues. The experimental results showed that the proposed method has higher segmentation accuracy compared with other methods.
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2009年第3期345-350,共6页
Chinese Journal of Biomedical Engineering
基金
国家自然科学基金资助项目(30770608)
上海市教委重点项目(06ZZ33)
上海市教委科研创新项目(09YZ216)
关键词
支持向量机
独立分量分析
特征优化
脑组织分割
support vector machine （SVM）
independent component analysis （ICA）
feature optimization
brain tissue segmentation
